<div class="gmail_quote">On Sun, Dec 25, 2011 at 21:37, Pavan Balaji <span dir="ltr"><<a href="mailto:balaji@mcs.anl.gov">balaji@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I meant predefined datatype, not OP. I believe the limitation in this particular case is not the operations, just the datatypes, anyway.</blockquote></div><br><div>Yes, exactly. Is there anything I can do to help move this forward?</div>
<div><br></div><div>We have had support for __float128 for about a year in PETSc and we can demonstrate applications where it is critical. I have some new parallel primitives that we would like to use more extensively in PETSc, but the most efficient (lowest setup cost) implementation relies on MPI_Accumulate().</div>